Clients SHOULD NOT include a
Referer
header field in a (non-secure) HTTP request if the referring page was transferred with a secure protocol.” htt
When you do a Google Search with Google Chrome, the following tag appears in the search results:
<meta content="origin" id="mref" name="referrer">
The origin
value means that instead of completely omitting the Referer
when going to http
from https
, the origin domain name should be provided, but not the exact page within the site (e.g. search strings will remain private).
On the other hand, link aggregators like lobsters have the following, which ensures that the whole URL will always
be provided in the Referer
(by browsers like Chrome and Safari), since link stories are public anyways:
<meta name="referrer" content="always" />
As of mid-2014, this meta[@name="referrer"]
is just a proposed functionality for HTML5, and it doesn't appear to have been implemented in Gecko, for example -- only Chrome and Safari are claimed to support it.
http://smerity.com/articles/2013/where_did_all_the_http_referrers_go.html
https://bugzilla.mozilla.org/show_bug.cgi?id=704320
http://wiki.whatwg.org/wiki/Meta_referrer
cnst answers this correctly above; it's content="origin". That forces browsers going HTTPS->HTTPS and HTTPS->HTTP to have the request header:
http-referer=https://www.google.com
This functionality allows sites to get credit for traffic without leaking URL parameters to a third party. It's awesome, as it's so much less hacky than what people have used here in the past.
There are currently three competing specs for this. I don't know which one is authoritative, and suspect it's a mix. They're similar, on most points.
Here's available support, that I know of; would love for people to let me know if I'm wrong or missing anything.
Now:
Unknown version:
Upcoming real soon now:
I think its because Google uses
<meta name="referrer" content="always">
So when a person goes from HTTPS to a HTTP site, the referrer is kept. Otherwise, without this the referrer would be stripped.